{
  "fetched_at": "2026-10-18T13:49:41Z",
  "source_url": "https://www.kaggle.com/competitions/cse151b-spring/data",
  "pages": [
    {
      "name": "data-description",
      "content": "<p>This page describes the details of the dataset. We provide the training set and the input of the test set. You are requested to submit the output (predictions) for the test set. The test set is further split into 80% for the public leaderboard and 20% for the private leaderboard.  </p>\n<p> We also provide a sample submission file with zero predictions. You are expected to replace the zeros with the predictions generated by your deep learning model. </p>\n\n<h2>File descriptions</h2>\n<ul>\n<li><strong>new_train</strong> - the training set including the output labels</li>\n<li><strong>new_val_in</strong> - the test set with the inputs only</li>\n<li><strong>sampleSubmission.csv</strong> - a sample submission file in the correct format</li>\n</ul>\n<h2>Data fields</h2>\n<p>Every pickle is a scene with trajectory tracking sequences for a maxium of 60 vehicles recorded at 10HZ (10 records per second). The target vehicle is called agent with a corresponding agent_id. Track_id is a sequence of IDs that track the ID of all vehicles including the agent itself.  The pickle file contains a dictionary with the following keys/values \n<ul>\n<li><strong>p_in</strong>-the (x,y) position input in the first two seconds (19 time steps) </li>\n<li><strong>v_in</strong>-the (x,y) velocity input in the first two seconds (19 time steps) </li>\n<li><strong>p_out</strong>- the (x,y) position output in the next three seconds (30 time steps)</li>\n<li><strong>v_out </strong>- the (x,y) velocity output in the next three seconds (30 time steps)</li>\n<li><strong>track_id</strong>- the track_id for each vehicle in the output sequence (30 time steps) </li>\n<li><strong>scene_idx</strong>- the id of this scene</li>\n<li><strong>agent_id</strong>- track id for the agent in this scene</li>\n<li><strong>car_mask</strong>- boolean index for the real car, we need to align the car numbers</li>\n<li><strong>lane</strong>- (x,y) for centerline nodes in this scene</li>\n<li><strong>lane_norm</strong>- the direction of each lane node</li>\n</ul>\n\n<h2>Submission Format</h2>\nYou will submit a csv file with 3,201 rows and 61 columns. The first row is the header indicating ID, v1, v2...\nThe first column corresponds to scene_idx. The rest of the 60 columns are p_out (x,y) over 30 time steps. ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "<h2>Background</h2>\n\n<p>Autonomous vehicles (AV)  are expected to dramatically redefine the future of transportation. However, there are still significant engineering challenges to be solved before one can fully realize the benefits of self-driving cars. One such challenge is building models that reliably predict the movement of traffic agents around the AV, such as cars, cyclists, and pedestrians.\n\n</p><p>\nIn this competition, you’ll apply your data science skills to build motion forecasting models for self-driving vehicles.  You'll have access to a large AV motion dataset to train and test your models. The goal of the motion forecasting task is to predict the positions of a tracked object 3 seconds into the future, given an initial 2-second observation. Your knowledge of deep learning will then be required to predict how cars  move in the AV's environment.\n</p>\n\n<p>\nThe training, validation, and test sequences are taken from different areas of Pittsburgh and Miami so that there is no geographical overlap. Each sequence follows the trajectory of the main agent for 5 seconds, while keeping track of all other agents including cars and pedestrians.   Agents follow very complex trajectories, such as changing lanes, navigating intersections, and turning.  We also provide the HD map information that is associated with the trajectory.\n</p>\n\n<h2>Team Guideline</h2>\nThis is a group project for the class.  You can request the computational resources from UCSD Data Science/Machine Learning platform allocated for this course. See more details on Course Website. \n\n- Maximum submission per day: 10\n- Submission deadline: 11:59 pm May 28, 2021\n\n\n\n\n\n<h2>Acknowledgements</h2>\n<p>We thank Argoverse for sharing this dataset.</p>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "<p>The evaluation pages describes how submissions will be scored and how you should format their submissions. </p>\n<hr>\n\n<p>We will use the following evaluation metrics for this competition, see more details at <a href=\"https://eval.ai/web/challenges/challenge-page/454/evaluation\"> Argoverse Evaluation Criteria </a>. \n\n-  Root Mean Squared Error  (RMSE): The average Root Mean Squared Error distance between the best-forecasted trajectory and the ground truth.  \n\n\n<h2>Submission Format</h2>\n<p><strong>For every scene in the dataset</strong>, submission files should contain 61 columns: scene ID and predicted x, y positions over 30 time steps. scene ID should be a string. Every x, y positions should be listed in a time-ordered fashion. That means x_1, y_1, x_2, y_2...</p>\n<p>The file should contain a header and have the following format:</p>\n<pre>ID, v1<br>1,1<br></pre>\n\nSee the sample_submission.csv file in the data folder for an example.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "This is a page where you can include rules that participants must accept before joining. You may wish to include rules like:\n\n - Don't cheat!\n - Apply yourself!\n - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}